How to transfer usb microphone streaming data at 44.1kHz

> (3) When I send 9 x 176 bytes and one 180 bytes every 10 millisecond.
This could be the right way,
if you would write it as "10 frames", instead of "10 milliseconds".

1 frame is not exactly 1 ms. It’s 1ms +/- 500ns, determined by the host.
In Synchronous method, your local sampling clock has to synchronize to SOF timing.

I recommend you [b]A[/b]synchronous instead of Synchronous, for easy microphone implementation.

In Asynchronous method, your local sampling clock is the master clock.
The number of samples in a packet is determined by the number of samples in the last SOF interval. And then, you don’t need to worry about either 44.1 or 48 kHz.

Tsuneo

Thanks for your reply.

I had solved this problem now. It is because of the Max packets size of IN_EP was set to 176 bytes.

But I am actually working at the SYNC problem(FIFO overrun & underrun). I am still learning on that three methods(sync, async, adaptive) and I think it is hard for me.

> I am still learning on that three methods(sync, async, adaptive) and I think it is hard for me.

The synchronization concept is not so difficult, though the implementation is another story.
As the USB audio consists a real-time system, host, bus and device should be synchronized each other. Sync, async and adaptive mean the place where the master clock exists.

Synchronous
- bus (ie. SOF timing) is the master clock. Host and device synchronize to SOF

Asynchronous
- Device has the master clock. Bus transfer and host synchronize to device

Adaptive
- Host has the master clock. Bus transfer and device synchronize to host.

Device declares Synchronization method at bmAttributes field (bit3..2) of the isoc endpoint descriptor.
00 = No Synchronization
01 = Asynchronous
10 = Adaptive
11 = Synchronous

A) Synchronous microphone
Here is a typical implementation of Synchronous microphone.

1) Generate local sampling clock, synchronized to SOF
A timer, driven by SysClock, counts SOF interval at every SOF timing (triggered by hardware SOF output or SOF interrupt). The count is divided by sample numbers at frames, and divided values are fed to another timer to generate sampling clock.

For example, 10 frames of SOF intervals counts up to 500,021 counts over 50 MHz SysClock.
500021 counts / (44.1 samples/frame * 10 frames) = 1133 base counts / sample .. 368 remainder

The second interval timer (sampling clock generator), also driven by SysClock, is set to 1134 counts (base+1) for the first 368 intervals, and 1133 base counts for the rest (441 - 368) intervals.

This operation is repeated at every 10 frames.
In this way, you have sampling clock, synchronized to SOF.

2) Sample data on a buffer
Driven by above sampling clock, an ADC fills sample data to a buffer.

3) Transfer to host
At every SOF timing, the increments of the samples on the buffer is captured. The firmware sends samples of the increments from the buffer to the isoc endpoint.

B) Async microphone
The implementation of Async microphone is almost same, except,
1') The device runs its own sampling clock.

The frequency of this sampling clock doesn't need to be exactly 44.1kHz,
because the bus transfer and host synchronize even to slightly-deviated sampling clock.

Tsuneo

Thanks for your reply again. It is very helpful to me. Thankyou very much!

Jason